update model

README.md

@@ -24,10 +24,10 @@ model-index:
     metrics:
        - name: Test WER
          type: wer
-         value: 11.83
+         value: 11.81
        - name: Test CER
          type: cer
-         value: 4.09
+         value: 4.06
 ---
 
 # Wav2Vec2-Large-XLSR-53-Portuguese
@@ -49,7 +49,7 @@ from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
 
 LANG_ID = "pt"
 MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-portuguese"
-SAMPLES = 5
+SAMPLES = 10
 
 test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
 
@@ -81,11 +81,16 @@ for i, predicted_sentence in enumerate(predicted_sentences):
 
 | Reference  | Prediction |
 | ------------- | ------------- |
-| NEM O RADAR NEM OS OUTROS INSTRUMENTOS DETECTARAM O BOMBARDEIRO STEALTH. | NENHUM VADA ME OS SALTOWS INSTRUMENTOS DE TECTERÁM UM BOMBADEIRO STER |
-| PEDIR DINHEIRO EMPRESTADO ÀS PESSOAS DA ALDEIA | EDIR DINHEIRO EMPRESTADO ÀS PESSOAS DO ALDEIRA |
+| NEM O RADAR NEM OS OUTROS INSTRUMENTOS DETECTARAM O BOMBARDEIRO STEALTH. | NEM UM VADAMEIM OS SOFTWART'S INSTRUMENTOS DE TETEN UM BAMBEDEIRO STER |
+| PEDIR DINHEIRO EMPRESTADO ÀS PESSOAS DA ALDEIA | PEDIR GINHEIRO E EMPRESTAR DAS PESSOAS DO ALDEIA |
 | OITO | OITO |
-| TRANCÁ-LOS | TRAM COULTS |
+| TRANCÁ-LOS | TRÃM CALMOS |
 | REALIZAR UMA INVESTIGAÇÃO PARA RESOLVER O PROBLEMA | REALIZAR UMA INVESTIGAÇÃO PARA RESOLVER O PROBLEMA |
+| O YOUTUBE AINDA É A MELHOR PLATAFORMA DE VÍDEOS. | YOUTUBE AINDA É A MELHOR PLATAFORMA DE VÍDEOS |
+| MENINA E MENINO BEIJANDO NAS SOMBRAS | MENINA E MENINO BEIJANDO NAS SOMBRAS |
+| EU SOU O SENHOR | EU SOU O SENHOR |
+| DUAS MULHERES QUE SENTAM-SE PARA BAIXO LENDO JORNAIS. | DUAS MERES QUES SENTAM-SE PARA BAIXO GRANDES DE UM NAIS |
+| EU ORIGINALMENTE ESPERAVA | EU ORIGINALMENTE ESPERAVA |
 
 ## Evaluation
 
@@ -102,9 +107,11 @@ LANG_ID = "pt"
 MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-portuguese"
 DEVICE = "cuda"
 
-CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞", 
+CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
                    "؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
-                   "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
+                   "{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
+                   "、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
+                   "『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "ʻ", "ˆ"]
 
 test_dataset = load_dataset("common_voice", LANG_ID, split="test")
 
@@ -156,7 +163,7 @@ In the table below I report the Word Error Rate (WER) and the Character Error Ra
 
 | Model | WER | CER |
 | ------------- | ------------- | ------------- |
-| jonatasgrosman/wav2vec2-large-xlsr-53-portuguese | **11.83%** | **4.09%** |
+| jonatasgrosman/wav2vec2-large-xlsr-53-portuguese | **11.81%** | **4.06%** |
 | joorock12/wav2vec2-large-xlsr-portuguese-a | 15.52% | 5.12% |
 | joorock12/wav2vec2-large-xlsr-portuguese | 15.95% | 5.31% |
 | gchhablani/wav2vec2-large-xlsr-pt | 17.64% | 6.04% |

config.json

@@ -7,6 +7,8 @@
   ],
   "attention_dropout": 0.1,
   "bos_token_id": 1,
+  "codevector_dim": 768,
+  "contrastive_logits_temperature": 0.1,
   "conv_bias": true,
   "conv_dim": [
     512,
@@ -37,12 +39,14 @@
   ],
   "ctc_loss_reduction": "mean",
   "ctc_zero_infinity": true,
+  "diversity_loss_weight": 0.1,
   "do_stable_layer_norm": true,
   "eos_token_id": 2,
   "feat_extract_activation": "gelu",
   "feat_extract_dropout": 0.0,
   "feat_extract_norm": "layer",
   "feat_proj_dropout": 0.05,
+  "feat_quantizer_dropout": 0.0,
   "final_dropout": 0.0,
   "gradient_checkpointing": true,
   "hidden_act": "gelu",
@@ -66,11 +70,15 @@
   "mask_time_selection": "static",
   "model_type": "wav2vec2",
   "num_attention_heads": 16,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
   "num_conv_pos_embedding_groups": 16,
   "num_conv_pos_embeddings": 128,
   "num_feat_extract_layers": 7,
   "num_hidden_layers": 24,
+  "num_negatives": 100,
   "pad_token_id": 0,
-  "transformers_version": "4.5.0.dev0",
+  "proj_codevector_dim": 768,
+  "transformers_version": "4.7.0.dev0",
   "vocab_size": 46
 }

preprocessor_config.json

@@ -1,5 +1,6 @@
 {
   "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
   "feature_size": 1,
   "padding_side": "right",
   "padding_value": 0.0,

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c8b135f54fc90d03a93c3cee21c655b9652acb2d54619af1520b0cb0d2f6a437
+oid sha256:2a34faf4721b64347b454d6b83095b764f25d6fc32df31908929e85a0080b476
 size 1262122455

vocab.json

@@ -1 +1 @@
-{"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "|": 4, "Ã": 5, "U": 6, "À": 7, "V": 8, "M": 9, "S": 10, "J": 11, "Ü": 12, "C": 13, "'": 14, "L": 15, "P": 16, "W": 17, "Y": 18, "I": 19, "É": 20, "F": 21, "Â": 22, "Z": 23, "D": 24, "H": 25, "R": 26, "K": 27, "Q": 28, "Á": 29, "B": 30, "-": 31, "Ç": 32, "Ê": 33, "E": 34, "G": 35, "T": 36, "O": 37, "N": 38, "X": 39, "Õ": 40, "Ô": 41, "Í": 42, "A": 43, "Ú": 44, "Ó": 45}
+{"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "|": 4, "'": 5, "-": 6, "A": 7, "B": 8, "C": 9, "D": 10, "E": 11, "F": 12, "G": 13, "H": 14, "I": 15, "J": 16, "K": 17, "L": 18, "M": 19, "N": 20, "O": 21, "P": 22, "Q": 23, "R": 24, "S": 25, "T": 26, "U": 27, "V": 28, "W": 29, "X": 30, "Y": 31, "Z": 32, "À": 33, "Á": 34, "Â": 35, "Ã": 36, "Ç": 37, "É": 38, "Ê": 39, "Í": 40, "Ó": 41, "Ô": 42, "Õ": 43, "Ú": 44, "Ü": 45}